Ribonuclease III Processing of Coaxially Stacked RNA Helices*

The RNase III family of endoribonucleases participates in maturation and decay of cellular and viral transcripts by processing of double-stranded RNA. RNase III degradation is inherent to most antisense RNA-regulated gene systems in Escherichia coli. In the hok/sok system from plasmid R1, Sok antisense RNA targets the hok mRNA for RNase III-mediated degradation. An intermediate in the pairing reaction between Sok RNA and hok mRNA forms a three-way junction. A complex between a chimeric antisense RNA andhok mRNA that mimics the three-way junction was cleaved by RNase III both in vivo and in vitro. Footprinting using E117A RNase III binding to partially complementary RNAs showed protection of the 13 base pairs of interstrand duplex and of the bottom part of the transcriptional terminator hairpin of the antisense RNA. This suggests that the 13 base pairs of RNA duplex are coaxially stacked on the antisense RNA terminator stem-loop and that each stem forms a monomer half-site, allowing symmetrical binding of the RNase III dimer. This processing scheme shows an unanticipated diversity in RNase III substrates and may have a more general implication for RNA metabolism.

The RNase III family of endoribonucleases participates in maturation and decay of cellular and viral transcripts by processing of double-stranded RNA. RNase III degradation is inherent to most antisense RNA-regulated gene systems in Escherichia coli. In the hok/sok system from plasmid R1, Sok antisense RNA targets the hok mRNA for RNase III-mediated degradation. An intermediate in the pairing reaction between Sok RNA and hok mRNA forms a three-way junction. A complex between a chimeric antisense RNA and hok mRNA that mimics the three-way junction was cleaved by RNase III both in vivo and in vitro. Footprinting using E117A RNase III binding to partially complementary RNAs showed protection of the 13 base pairs of interstrand duplex and of the bottom part of the transcriptional terminator hairpin of the antisense RNA. This suggests that the 13 base pairs of RNA duplex are coaxially stacked on the antisense RNA terminator stem-loop and that each stem forms a monomer half-site, allowing symmetrical binding of the RNase III dimer. This processing scheme shows an unanticipated diversity in RNase III substrates and may have a more general implication for RNA metabolism.
RNase III of Escherichia coli (1) belongs to a family of double strand-specific endoribonucleases (2,3), which have retained interspecies substrate cleavage specificity (4). This family of enzymes comprises both prokaryotic and eukaryotic members whose primary function is processing of ribosomal RNA precursors (4 -6). However, substrates also include several mRNAs from E. coli, bacteriophages and T7 (2,3), U5 small nuclear RNA (7), U2 small nuclear RNA (8), and a large subset of small nucleolar RNA precursors of Saccharomyces cerevisiae (9,10). Substrate recognition requires two helical turns of an A-form RNA helix, and processing occurs with a consensus 2-nucleotide, 3Ј-recessive, staggered cut creating 5Ј-phosphate and 3Ј-hydroxyl termini (11,12). The E. coli RNase III holoenzyme is a homodimer of 52 kDa that binds dsRNA 1 in a quasisymmetrical fashion concentrical of the scissile phosphodiester bonds (13). RNase III binding is dependent on a conserved dsRNA-binding motif present in the C-terminal end that forms an ␣-␤-␤-␤-␣ fold (14). Binding of dsRNA is believed to occur primarily via non-electrostatic interactions with a helical ar-rangement of 2Ј-hydroxyl groups in the minor groove as described for the dsRNA-binding motif binding of the mammalian dsRNA-activated protein kinase, PKR (15). In addition, a lysine residue may be required for dsRNA binding, similar to what has been observed for the staufen protein from Drosophila (16) and PKR (17)(18)(19). Binding of the dsRBD of RNA-binding protein A from Xenopus laevis to dsRNA has shown that two regions of the dsRBD contact successive minor grooves on the same face of the double-stranded RNA helix, whereas a third region contacts the spanning major groove (20). However, the recent finding of binding sequence anti-determinants in dsRNA that abolish RNase III cleavage in vitro (21) emphasizes the subtleties of substrate recognition by RNase III and may explain why certain dsRNA sequences like the human immunodeficiency virus type 1 TAR RNA hairpin are inherently refractory to processing (4).
Complete or partial duplexes of Ն20 bp formed between antisense RNAs and their target RNAs are efficiently processed by RNase III (22)(23)(24)(25)(26). The hok (host cell killing) mRNA from the hok/sok (suppression of killing) system of plasmid R1 (27) exists in three forms with alternative configurations and translational capacities (28). Sok antisense RNA consists of an 11-nucleotide 5Ј-single-stranded tail that is responsible for the initial interaction with hok mRNA and a hairpin that functions as a Rho-independent transcriptional terminator (Fig. 1A). Sok RNA represses hok translation by forming a 63-bp duplex with hok mRNA (Fig. 1C). The duplex is cleaved by RNase III in vivo and in vitro (26). Since formation of RNA/RNA binding intermediates could be faster than the formation of a full duplex between Sok RNA and hok mRNA, inactivation of the target RNA could occur prior to complete duplex formation, as in the cases of several other antisense RNA-regulated gene systems (29). To test this, we constructed a chimeric antisense RNA (CA-RNA) consisting of the 13-nucleotide 5Ј-tail of Sok fused to the terminator hairpin of PndB, an antisense RNA from plasmid R483 that is homologous to Sok (Fig. 1B). CA-RNA forms a 13-base pair duplex with hok mRNA, thereby resulting in the generation of a three-way junction (Fig. 1D, left panel). This complex mimics the naturally occurring binding intermediate between hok mRNA and Sok antisense RNA (Fig. 1D, right  panel). Surprisingly, CA-RNA was able to inhibit hok expression in vivo, 2 thus questioning the actual mechanism of hok target inactivation by Sok RNA.
Here, we have compared the effect of Sok RNA and CA-RNA on hok mRNA metabolism in rnc ϩ and rnc Ϫ strains. Complexes between antisense and target RNAs that correspond to the RNA complexes observed in vivo were assayed for RNase III binding and processing in vitro. Intriguingly, all data support a model in which RNase III recognizes and cleaves coaxially stacked RNA helices.
Construction of Plasmids-The hok Ϫ /sok ϩ and hok Ϫ /sok Ϫ constructs described previously (32) were inserted in the BamHI-EcoRI restriction sites of the low-copy number mini-R1 test plasmid pMH82 (Kan R ). Thus, pTF821(hok Ϫ /sok ϩ ) carries a hok/sok system with an amber codon in the hok gene, whereas pTF821(hok Ϫ /sok Ϫ ) carries an additional point mutation that inactivates the sok promoter. These mutations do not influence the stability or processing patterns of hok mRNA (30).
Northern Transfer Analysis-Northern analysis was carried out as described (26) using the W3110 (rnc ϩ ) and HT115 (rnc Ϫ ) E. coli strains grown in LB broth supplemented with 50 g/ml kanamycin and 100 g/ml ampicillin for retainment of the mini-R1 test plasmid and pBR322 antisense RNA-donating plasmid, respectively. 10 g/ml tetracycline was added to the HT115 strain.
Lead(II) Acetate Probing-5Ј-End-labeled CA-RNA or hok38 RNA was incubated either alone or in complex with an excess (1 pmol) of partner RNA in a buffer (50 mM Hepes-KOH (pH 7.5), 10 mM MgCl 2 , and 50 mM KCl) supplemented with 5 g of tRNA in a total reaction volume of 10 l. Lead(II) acetate (2.5 l) dissolved in H 2 O immediately prior to use was added to a final concentration of 0, 5, or 10 mM and incubated for 5 min at room temperature. Reactions were quenched by addition of EDTA to a final concentration of 40 mM. The RNA was precipitated, washed twice, resuspended in formamide dye, and subsequently resolved on 15% polyacrylamide gels containing 7 M urea and 1ϫ Tris borate/EDTA. RNA Cleavage Assay-5 nM RNA was incubated in the presence or absence of an 10-fold excess of various complementary partner RNAs in 1ϫ TMK-glutamate buffer (20 mM Tris acetate, 10 mM magnesium acetate, and 200 mM potassium glutamate) supplemented with 1 mM dithiothreitol and 1 g of tRNA in a reaction volume of 20 l. 50 nM N-terminal His 6 -tagged RNase III was added, and samples were withdrawn to formamide dye on ice at the time points indicated and subsequently loaded on 5.5 or 15% polyacrylamide gels containing 7 M urea and 0.5ϫ Tris borate/EDTA.
The E117A amino acid substitution was introduced by double polymerase chain reaction using the above primers and rnc-E117A-1 (5Ј-CCGACACCGTCGCAGCATTAATTGG-3Ј) and rnc-E117A-2 (5Ј-CCAATTAATGCTGCGACGGTGTCGG-3Ј), generating pTF601. LacI was expressed from the co-resident plasmid pMS421, which carries lacI q and a gene conferring spectinomycin resistance (34). Initial affinity purification was accomplished by protein binding to TALON TM resin in 1ϫ binding buffer (4% (NH 4 ) 2 SO 4 and 30 mM Tris-HCl (pH 8.0)). Three wash steps in 1ϫ binding buffer were followed by a wash in 1ϫ binding buffer supplemented with 5 mM imidazole as recommended by the manufacturer (CLONTECH). Proteins were eluted in 1 M NH 4 Cl with 150 mM imidazole and loaded directly onto a pre-equilibrated poly(I)⅐poly(C) AG column (Amersham Pharmacia Biotech) as described (33). Eluted protein was dialyzed against 1ϫ TMK-glutamate buffer supplemented with 1 mM dithiothreitol.
E117A RNase III Footprinting-One nM 5Ј-32 P-labeled CA-RNA alone or in complex with 10 nM 3 H-labeled hok13 was incubated with 0, 1, or 2.5 M E117A RNase III in 1ϫ TMK-glutamate buffer supplemented with 1 mM dithiothreitol and 5 g of tRNA in a 20-l reaction volume at room temperature. 0.1 unit of cobra venom nuclease (Amersham Pharmacia Biotech) was added for 5 min, and reactions were quenched by addition of 80 l of phenol. RNA samples were precipitated, washed, resuspended in formamide dye, and resolved on a 15% polyacrylamide gel containing 7 M urea and 0.5ϫ Tris borate/EDTA.
Cleavage Kinetics and Substrate Affinity Determinations-K m and k cat calculations were made by a double-reciprocal plot of 1/V (initial velocity) versus 1/S (substrate concentration) as described (33). 7.5 nM N-terminal His 6 -tagged RNase III dimer was incubated with 62-1000 nM preformed complexes between 32 P-labeled antisense RNA and equimolar 3 H-labeled target RNA in a 10-l reaction volume using TMK-glutamate buffer supplemented with 1 mM dithiothreitol (see above). Samples were quenched after 2 min by addition of formamide dye loading buffer. RNAs were resolved on a 15% polyacrylamide gel containing 7 M urea and 0.5ϫ Tris borate/EDTA.

RESULTS
In Vivo RNase III Processing of hok mRNA-To test the effect of natural and artificial antisense RNAs on the level of hok mRNA in vivo, a hok/sok system with abolished Sok RNA expression was constructed and cloned into a low-copy number mini-R1 plasmid (pTF821). To avoid detrimental expression of the Hok toxin, this construct carries an amber stop codon mutation in the hok gene (hok Ϫ ). Northern analysis showed that wild-type E. coli cells (strain W3110, rnc ϩ ) carrying pTF821 contain three hok mRNA species (Fig. 2, left panel, lane 2). We have shown previously that hok mRNA-1 and -2 are full-length mRNAs that are translationally inactive and bind Sok RNA inefficiently, whereas the third RNA is a 3Ј-truncated species (denoted Tr. in Fig. 2) that is translationally active and binds Sok RNA avidly (28). hok mRNA-2 is an RNase III cleavage product of RNA-1 (Fig. 2, compare left and right panels). This latter cleavage removes the 3Ј-terminal hairpin of RNA-1 and does not influence the analyses presented here (26).
Plasmid pTF820 is the isogenic sok ϩ derivative of pTF821 that produces wild-type Sok RNA in cis. Wild-type cells carrying pTF820 do not contain detectable amounts of the 3Ј-truncated mRNA due to its rapid interaction with Sok RNA and subsequent rapid cleavage by RNase III (Fig. 2, left panel, lane 1) (26). Consistent with this interpretation, the truncated RNA appears in cells devoid of RNase III (Fig. 2, right panel, lane 1).
We examined the effect of three antisense RNAs expressed in trans from pBR322-derived plasmids on the hok mRNA band pattern in rnc ϩ and rnc Ϫ cells (Fig. 2). The pBR322 control plasmid did not affect the hok mRNA band pattern (Fig. 2, left  panel, lane 3). In trans expression of wild-type Sok RNA from pTF322 resulted in the disappearance of the truncated mRNA (Fig. 2, left panel, lane 4). The high level of Sok RNA production also resulted in a modest reduction in hok mRNA-1 and -2, presumably due to antisense RNA binding to hok mRNA during its synthesis (Fig. 2, left panel, lane 4). Interestingly, expression of CA-RNA also caused the disappearance of truncated hok mRNA (Fig. 2, left panel, lane 5), an effect similar to that of Sok RNA. In contrast, wild-type PndB antisense RNA (pTF324) from plasmid R483 had no effect on hok mRNA (Fig.  2, left panel, lane 6).
When we examined the hok mRNA in an rnc Ϫ strain, neither Sok RNA nor CA-RNA affected the metabolism of truncated hok mRNA (Fig. 2, right panel). Thus, CA-RNA, which carries 13 nucleotides that are complementary to hok mRNA, is sufficient to target the truncated hok mRNA for degradation in the rnc ϩ strain. These results indicate that CA-RNA and the truncated hok mRNA form a partial duplex (i.e. the TWJ; see below) that is adequate for RNase III-mediated hydrolysis in vivo.
The CA-RNA⅐hok mRNA Complex Forms a Three-way Junction in Vitro-hok38 RNA constitutes the antisense target stem-loop structure in hok mRNA shown schematically in Fig.  1D. We tested the in vitro complex formed between CA-RNA and hok38 RNA by lead(II) acetate probing of 5Ј-end-labeled CA-RNA (Fig. 3A, left panel) or hok38 (Fig. 3A, right panel) either alone or in complex with hok38 or CA-RNA, respectively. The interpretation of the RNA probing is shown in Fig. 3B. Single strand-specific lead(II) acetate cleavages were observed in the 5Ј-tail and in the top stem and loop region of the unpaired CA-RNA. Binding of hok38 to 32 P-labeled CA-RNA resulted in repression of cleavages at the tail nucleotides only, whereas cleavages at other sites were unaffected (Fig. 3A, left  panel). Probing of unpaired hok38 resulted in cleavages in the loop and in the top stem region (Fig. 3A, right panel). Binding of CA-RNA specifically protected the nucleotides that are complementary to the bases in the tail of CA-RNA. Although binding of CA-RNA caused a modest enhancement of cleavages in the bottom stem of hok38 (Fig. 3A, right part of the left panel), the probing supports the formation of the three-way junction, as shown in Fig. 3B.
In Vitro RNase III Processing of the TWJ-Different antisense RNA/target RNA combinations were tested for their ability to function as RNase III-processing signals in vitro using purified histidine-tagged RNase III (Fig. 4). The cleavage reactions were performed at high stringency (i.e. at a high concentration of monovalent salt). As expected, truncated hok mRNA alone was not affected by RNase III (Fig. 4A, left panel), consistent with the extraordinary stability of this mRNA in vivo (28). Addition of Sok RNA resulted in a major cleavage product in addition to multiple consecutive cleavage fragments (MCF; Fig. 4A, middle panel). In contrast, addition of CA-RNA to truncated hok mRNA yielded two well defined cleavage products, indicative of a single specific cleavage site. Consistently, the longest cleavage fragments in the two reactions were of similar sizes (Fig. 4A).
We also examined RNase III cleavage of labeled wild-type Sok RNA (Fig. 4B) and CA-RNA (Fig. 4C). Alone, both RNAs were resistant to RNase III processing. The lack of cleavage is consistent with the presence of upper stem helix irregularities that may act to protect the RNAs from RNase III cleavage. Such stem irregularities are known to prevent RNase III cleavage (35). Addition of truncated hok mRNA resulted in multiple cleavages of Sok RNA (Fig. 4B, middle panel) and a single cleavage of CA-RNA (Fig. 4C, middle panel), consistent with the cleavage pattern observed for hok mRNA in Fig. 4A. These results show that all RNase III cleavages described here reflect coordinated double-strand scissions. Furthermore, primer-extension analysis showed that all double-strand cleavages occurred at the phosphodiester bond between nucleotides ϩ13 and ϩ14 of the antisense 5Ј-tail and the corresponding phosphodiester bond in hok mRNA, resulting in a 2-nucleotide 3Ј-overhang according to consensus RNase III processing (data not shown). This cleavage site is designated position 13/11 below.
As implied by the in vivo and in vitro RNase III processing experiments, a partial duplex between the CA-RNA antisense RNA and the hok mRNA seems to be sufficient for RNase III processing. A 14-nucleotide fragment (hok13) of hok mRNA sequence that is complementary to the 13-nucleotide 5Ј-tail of The hok mRNAs were expressed from a mini-R1 plasmid carrying either hok Ϫ /sok ϩ (pTF820) (lane 1) or hok Ϫ /sok Ϫ (pTF821) (lanes 2-6) gene systems. The hok Ϫ and sok Ϫ genotypes denote an amber mutation in the hok gene preventing toxin expression and a mutation in the sok promoter that abolishes Sok RNA expression, respectively (32). Lane 3, the pBR322 control plasmid; lanes 4 -6, wild-type Sok RNA (pTF322) and CA-RNA (pTF323) and PndB (pTF324) antisense RNAs produced in trans from pBR322-derived plasmids, respectively. Translationally inactive hok mRNA-1 and -2 as well as the active truncated (Tr.) hok mRNA are indicated. hok mRNA-2 is not produced in the rnc Ϫ strain.
Sok RNA and CA-RNA was synthesized (the fragment carried an additional 5Ј-G for efficient transcription by T7 RNA polymerase). Addition of hok13 to Sok RNA or CA-RNA resulted in single specific RNase III cleavages (Fig. 4, B and C, right panels) at the same processing site at position 13/11 described above (data not shown). These observations are consistent with the in vivo processing data and confirm that a partial antisense RNA/target RNA duplex comprising 13 bp of interstrand pairing is adequate for RNase III processing.
To test if all 13 bp of the duplex were required for processing, we examined 3Ј-end-shortened hok mRNA fragments of 7 (hok7) and 10 (hok10) nucleotides for RNase III-mediated CA-RNA cleavage (Fig. 4D). hok7 failed to sustain cleavage (Fig.  4D, left panel), whereas partial cleavage was observed with hok10 (middle panel). Thus, all 13 base pairs are required for optimal enzyme binding or processing. In addition, the cleavage observed with hok10 occurred at the site at position 13/11 described above, implying that binding and/or cleavage is fixed at a unique position and does not change with the 5Ј-end border of the partial duplex (data not shown).
Cleavage Kinetics and Substrate Affinity-The kinetic parameters for RNase III processing of Sok RNA in duplex with hok13 and CA-RNA in duplexes with hok13 and hok38 are shown in Table I. The k cat and K m values for RNase III processing of all three complexes are similar to the values calculated for the R1.1 substrate from bacteriophage T7 using wildtype RNase III (33). In addition, the formation of a TWJ in the CA-RNA⅐hok38 complex does not impair affinity or cleavage rate compared with the Sok RNA⅐hok13 and CA-RNA⅐hok13 complexes. Thus, the in vitro kinetic parameters for RNase III processing of the partial duplexes clearly support the ability of this substrate to compete for RNase III binding and cleavage in vivo.
The Transcriptional Terminator Stem-Loop Structure Is Prerequisite for RNase III Cleavage of hok13-Complexes formed between uniformly 32 P-labeled hok13 and Sok RNA, CA-RNA, or Sok13, which corresponds to the 13 nucleotides of the 5Ј-tail sequence of Sok RNA (and CA-RNA) were examined for accurate RNase III cleavage (Fig. 5). Binding of either Sok RNA or CA-RNA resulted in RNase III cleavage at the specific site in hok13. In contrast, RNase III failed to perform accurate processing of the hok13⅐Sok13 complex. Hence, the 13 base pair interstrand RNA duplex is not a substrate for RNase III. These results show that the transcriptional terminator hairpins of Sok RNA and CA-RNA are necessary for RNase III processing. The cleavage sites at position 13/11 suggest that the 13 bp of antisense RNA/target RNA duplexes and the bottom stem of the antisense RNA transcriptional terminator hairpins each constitute RNase III monomer half-sites.
RNase III Binding to Coaxially Stacked RNA Helices-RNase III binding was examined using an RNase III derivative carrying an E117A amino acid substitution. This protein shows binding properties identical to those of the wild-type enzyme, but is impaired in substrate cleavage (13). 32 P-5Ј-End-labeled CA-RNA was incubated either alone or with hok13 in the presence or absence of E117A RNase III (Fig. 6A). As the substrates tested contained almost exclusively double-stranded or stacked nucleotides, we used cobra venom nuclease to monitor substrate binding (Fig. 6B). At both concentrations of E117A RNase III, the unpaired CA-RNA was not protected from cobra venom nuclease cleavages. Adding the hok13 fragment resulted in E117A RNase III-mediated protection of the 5Ј-tail and bottom hairpin at both the 5Ј-and 3Ј-sides. These data are consistent with a symmetrical binding of E117A RNase III coaxially stacked helices that comprise the bottom part of the CA-RNA transcriptional terminator hairpin stacked on the 13 bp of the interstrand CA-RNA/hok13 duplex. A structural model showing the protection of the CA-RNA⅐hok13 complex by E117A RNase III is presented in Fig. 6B. Protection extends roughly 10 -12 bp on each side of the cleavage sites, which are located in the hinge region of the stacked helices (Fig. 6A). DISCUSSION Many antisense RNAs target their complementary transcripts for RNase III-mediated degradation. Although the antisense RNAs of the Sok RNA family can form complete duplexes with their target RNAs, our data show that a pairing intermediate of 13 base pairs that forms a TWJ is sufficient for RNase III-mediated degradation in vivo. Intriguingly, substrates corresponding to the transcripts examined in vivo were cleaved by purified RNase III at a single specific site in both strands of the antisense RNA/target RNA duplexes. Our data support a minimal substrate that consists of the antisense RNA transcriptional terminator hairpin and the 13 nucleotides of the 5Ј-tail engaged in RNA interstrand pairing. However, RNase III processing normally requires at least two helical turns of RNA duplex (2,3). Indeed, the 13 bp of duplex are necessary, yet insufficient for RNase III cleavage (Fig. 4). E117A RNase III protein footprinting supports enzyme binding at the 13 bp of interstrand duplex and at the bottom 10 -12 bp  FIG. 5. In vitro RNase III cleavage of uniformly 32 P-labeled hok13 transcripts. The assays were conducted as described in the legend to Fig. 3 at 37°C. The alkali ladder (L) was made on a 32 P-5Јend-labeled hok13 fragment as described (36). Uncleaved RNA (UC) and the 5Ј-and 3Ј-cleavage products corresponding to 3 and 11 nucleotides, respectively, are indicated. of the antisense RNA hairpin stem. This substrate most likely forms a rigid coaxially stacked structure that constitutes more than two helical turns of RNA helix, thus being reminiscent of more conventional RNase III substrates. The observation that RNase III is unable to perform any cleavage on the 13-bp duplex substrate suggests either of the following. (i) Simultaneous contact between the substrate half-sites and the dsRNAbinding motifs of each RNase III subunit is required to confer catalytic activity to the enzyme. This could occur by structural transitions of the enzyme upon substrate binding, perhaps to attain the transition state. (ii) Binding of a single monomer subunit to the RNA complex provides insufficient binding energy to sustain substrate cleavage.
Consistent with the above-described conclusions, the footprinting experiment showed that binding of the dimeric E117A RNase III requires both monomer half-sites of the CA-RNA⅐hok13 complex (Fig. 6). The absence of protection of the monomer half-site in CA-RNA suggests that the equilibrium dissociation constant for the interaction between the RNase III dsRBD and the half-site is significantly higher than 2.5 M. A low binding affinity for such a monomer/half-site interaction is expected to increase the stringency in substrate binding and to prevent titration of RNase III by dsRNA elements that are not RNase III substrates. The K m and k cat values for CA-RNA in partial duplex with hok13 and hok38 and for Sok RNA in complex with hok13 are similar to the values previously reported for RNase III processing (33) and concur with the observation that the CA-RNA⅐hok mRNA substrate is efficiently processed in vivo. Thus, formation of the TWJ in the CA-RNA⅐hok mRNA complex (Figs. 1D and 3) does not impede RNase III processing (Figs. 2 and 4 and Table I), which could suggest that the enzyme does not form a tightly closed complex surrounding the cleavage site since the extra RNA stem is accommodated by the catalytic site without affecting k cat .
Several RNase III substrates have structural irregularities (i.e. bulges, mismatches, and internal loops) (2) around the scissile bonds, including the Sok RNA⅐hok13 and CA-RNA⅐hok13 complexes and the R1.1 substrate of bacteriophage T7. However, data obtained with the native R1.1 substrate of bacteriophage T7 and an engineered version with perfect helicity showed similar equilibrium dissociation constants, suggesting that little or no discrimination on substrate binding is conferred by the catalytic site (13). Instead, the recent finding of RNase III binding anti-determinants implies that the stringency of processing is on the level of enzyme binding exclusively (21). Consistent with the efficient processing, none of the substrates tested here convey the RNase III binding anti-determinants. Alignment of RNase III-processing signals from E. coli has identified a preference for a C-G base pair at position ϩ6/ϩ4 relative to the scissile bonds (21,37). Interestingly, the CA-RNA⅐hok mRNA complex contains one C-G base pair in the tail/target duplex and one C-G base pair in the terminator hairpin of CA-RNA located symmetrically (disregarding the bulged-out U) at position ϩ6/ϩ4 (Fig. 6B). The crystal structure of the dsRBD of the dsRNA-binding protein A of X. laevis in complex with dsRNA revealed a single sequence-specific contact between a backbone carbonyl group and the exocyclic amine of G in the minor groove of a C-G base pair (20). Thus, a similar interaction could be important for RNase III binding of its substrates and may in some cases specify the cleavage site location. This is supported by the fixed cleavage site at position 13/11 observed here for RNase III processing of the CA-RNA⅐hok10 complex, for which the end of the interstrand duplex has been juxtaposed, compared with the CA-RNA⅐hok mRNA and CA-RNA⅐hok13 complexes.
The fact that the family of RNase III enzymes shows at least partial interspecies substrate specificity (4) suggests that coaxially stacked helices could function as processing signals for RNase III enzymes in both prokaryotic and eukaryotic cells. Recently, an RNase III-processing signal composed of noncontiguous helices was suggested for the small nuclear R40 precursor of S. cerevisiae (10). However, for this substrate, the junction of the RNA helices is positioned 6 -8 base pairs from the scissile phosphodiester bonds, most likely displaced from the catalytic site.
Intriguingly, binding of coaxially stacked dsRNA by two copies of the double dsRBD may apply to a diverse set of proteins with unrelated function. In a recent experiment, the dsRBD of PKR was shown to bind in vitro selected RNA with noncontiguous and most likely coaxially stacked RNA helices (38). Thus, it is conceivable that enzymes like PKR and ADAR (adenine deaminase that acts on RNA) that carry multiple copies of the dsRBD can bind coaxially stacked RNA stems in a fashion similar to RNase III and thereby trigger such diverse cellular responses as apoptosis, the interferon-induced viral response (39,40), or RNA degradation (41).  3 and 7) and probed with cobra venom ribonuclease (lanes 1-3 and 5-7). The product(s) of cleavage reactions using wild-type RNase III (lanes 4 and 8) are shown by an arrowhead. RNase T1 cleavage (T1) on native RNA and the alkali ladder (L) were as described (36). C indicates the control reaction in which E117A RNase III and cobra venom nuclease were omitted. B, the secondary structure and coaxially stacked helices of the CA-RNA⅐hok13 complex. The hok/sok sequence is shown in boldface, and the PndB sequence in lightface. The approximate extent of E117A RNase III protection from A is shown with boldface lines. The proposed symmetrical binding of RNase III on each half-site is illustrated by double barrels, and the positions of the staggered dsRNA breaks are indicated by arrows. The symmetrically located C-G base pairs described under "Discussion" are boxed.
The processing of coaxially stacked RNA helices reported here shows that RNase III substrates are more versatile than previously assumed. This novel substrate specificity could have a more general implication on the metabolism of diverse RNAs.